Leveraging Clustering Techniques to Facilitate Metagenomic Analysis

Machine learning clustering algorithms provide excellent methods for conducting metagenomic analysis with efficiency. This study uses two machine learning algorithms, the self-organizing map and the K-means algorithms, to cluster data from an environmental sample collected from a hot springs habitat and to provide a visual analysis of that data. A data processing pipeline is described that uses the clustering algorithms to identify which reference genomes should be included for further analysis in determining possible organisms that are present in a metagenomic sample. The clustering revealed probable candidates for additional analysis, including a thermophilic, anaerobic bacterium, which is likely to be found in a hot springs environment and serves to validate the functionality of these tools. The machine learning techniques discussed here can serve as a launching point for elucidating protein sequences that could serve as possible reference comparisons to a specific metagenomic sample and lead to further study.